View Full Version : Huge rendering problem!!! HELP!!!

10-13-2003, 03:29 PM
I am using frenchwave's split render to try and output a rendering 5080 pixels high by 2050 pixels high for a 7foot high banner. No matter how I split the file with splitrender; 6 x 6 or 10 x 10, it won't render. I have tried it on the following machines.

A P4 with a gig of ram, virtual memory maxed.
A dual PIII 500 with 500 megs of ram, virtual mem maxed.

Both machines won't render them, If i use LWSN controller, I get no errors, it just frees the buffers and moves on, if I try rendering each generated scene through lightwave, I get memory errors.

I thought the split render was supposed to alleviate these issues. From reading other threads, others seem to have success with this.

Any help or input would be greatly appreciated!

Facial Deluxe
10-13-2003, 03:46 PM
Try to join FrenchW at

10-13-2003, 04:00 PM
Thanks Facial Deluxe
That was interesting trying to register with that forum. Thank god I'm familiar with forum layouts or i would have been lost since everything is in french.

Pulling my hair out. Got this nice bald spot on the side of my head.

10-13-2003, 08:08 PM
I've setup lightnet to take advantage of both machines at once. Still seem to be having memory issues. I turned down the memory segments to like 250mb, and resplit the scene. Still have memory issues though. Running close to deadline now and not sure what to do. Is there something better on the market to use?

10-13-2003, 11:21 PM
If you're having memory issues, perhaps lowering your sub-d render level would help.

You can go as low as 1 in some cases, depending on how far the object is from the camera...

10-13-2003, 11:40 PM
Does that make any difference if I have no sub-d'd objects in my scene?

I reduced my segment memory from 750 to 250 and was able to start the rendering with an F9, but still having problems running it through lightnet, or screamernet.


10-14-2003, 08:36 AM
Maybe I understand you wrong but if the LWSN controller works why don't you split the image manually and put it together in Photoshop or what else image program. What I mean is to set up a scenefile for every segment you what to render, create a batch file for every computer and start rendering from a command prompt. If you split the image in for example 10 segments you can render, depending on the speed of the machines and the scene complexity, segment 1 to 7 on your P4 and segment 8 to 10 on the PIII.

You can use the notepad to setup the segments. Just open your scene file in the notepad and search for the following to lines:

First line is the segment:
'RegionLimits 0 1 0 1'

To setup 10 'row' segments change it to the following:
'RegionLimits 0 1 0 0.1'

Second line is the output image
'SaveRGBImagesPrefix Q:\output\testimage\monsterrender'

Give the image a number to avoid overwriting:
'SaveRGBImagesPrefix Q:\output\testimage\monsterrender_01_'

Now save your scene as 'myscene_seg01.lws'

for the second segment change the lines to:
'RegionLimits 0 1 0.1 0.2'
'SaveRGBImagesPrefix Q:\output\testimage\monsterrender_02_'

and save the scene as 'myscene_seg02.lws'

Go on so until:
'RegionLimits 0 1 0.9 1'
'SaveRGBImagesPrefix Q:\output\testimage\monsterrender_10_'

Be carefully if you want to put out more images (for example special Buffers) that you also give them numbers.

Now you can create a batch file for every computer. For example:

lwsn -3 scenes\myscene_seg01.lws 1 1 1
lwsn -3 scenes\myscene_seg02.lws 1 1 1
lwsn -3 scenes\myscene_seg03.lws 1 1 1
lwsn -3 scenes\myscene_seg04.lws 1 1 1
lwsn -3 scenes\myscene_seg07.lws 1 1 1

for the P4 and

lwsn -3 scenes\myscene_seg08.lws 1 1 1
lwsn -3 scenes\myscene_seg09.lws 1 1 1
lwsn -3 scenes\myscene_seg10.lws 1 1 1

for the PIII

Sorry if this dosen't make sense in your situation because if I understand you wrong.
BTW, I think LWSN works because it don't use the segment memory settings of the layouter or am I wrong?



10-14-2003, 04:46 PM

I have the scene rendering in layout on my dual pentium by dropping the segment memory down to 100mb. I appreciate your thoughts on the problem and am going to run some tests with that on a seperate machine. I'm scared to stop the rendering at this point since it's about 15 hours into it now. I need to build a reliable render farm.

I'll let you know how i make out with your suggestions.

10-14-2003, 10:52 PM
Originally posted by carbon8
Does that make any difference if I have no sub-d'd objects in my scene?

Yes it does. Higher sub-d levels (like > 2) can really consume mass quantities of ram at render time. And if you're trying to render with two screamernet nodes either through LW or lightnet, you;re going to need twice as much ram to do it.

If your models are Sub-D, but don't deform, it would be better to freeze them and render those versions.

But also, if your scene is poly-heavy, there may not be any way around it, other than u sing a plugin like Microwave from Evasion to lower your poly count.

At least that's been my experience.

10-14-2003, 11:12 PM
Thanks lfgabel
I'll keep that in mind. Good info!

Poly count is 21,350 with sub-d's set to 3.
Here's a test render I did last week before tweeking the lights and transparency maps a little.


I can't get lightnet or screamernet to work. They just hang. I have everything mapped properly, but think it's my wireless network. Too slow (802b). Has anyone else had problems on a wireless network.????????????

W Wade
10-15-2003, 12:00 AM
The size of your image / transparency maps may have a lot to do with your troubles. If you can down size some of them it will help. I have had problems with one render as my 25 maps totaled 35+Megs and LW would crash on such a large render. After reducing the size of my maps to a sum total of 8 Megs it would render out just fine..

Looks like reducing map size may be a hard thing for you to do thou; as it looks as if there are quiet a bit of them and that they are rather detailed.

10-15-2003, 12:07 AM
Will reducing the size of the maps hurt the detail on the final render which will be printed out at roughly 7 feet tall?
If not, I can run a batch in photoshop to reduce them. Some of them are roughly 10megs.

W Wade
10-15-2003, 03:03 AM
I would try reducing the size of each map to 36dpi because if you have a 10 Meg (300dpi) or so image map for each bag of chips you will have 300megs in images for just your potato chips! Do you really need that much detail? Is your finale rendering going to be 300 ++ Megs? I would guess if you reduce maps down to less than 2 Megs each you would find you have enough detail to get you the image you want.

I wonder if a person could render an image with no maps and see if that is the problem. Seems like a hassles but it would key you into whether or not that is the hang up.

Try this – Save all you textures into your preset folder you can do this by running a small f9 render with viper activated double click on the texture in the small window and it will save into your preset folder – a really nice feature this is. Then resize all you image maps and rename them something like chips at 36dpi… or whatever. Now replace each image map with the much smaller size AND RENAME each texture with the 36dpi at the end or whatever. I really think it is an image map problem and for an hours work or less you should be able to find out. If having your image maps at 36 dpi works and looks good your good to go. If your render looks a little blurry size up your maps to 100 – 200 or whatever dpi gives you an excitable image.

10-15-2003, 04:33 AM
carbon8, this image by werner was posted on this forum
a few months ago. I saw the same image at siggraph and it was printed about 10 feet high. Not sure what his resolution was, but the 10 foot image looked pretty sharp.

W Wade
10-15-2003, 05:16 AM
I think you can get a high res render no problem it just the high res image maps that you throw in the mix that will kill the deal.

I maybe wrong but only if I am not right. :D

10-15-2003, 07:42 AM
Originally posted by carbon8
Thanks lfgabel
I'll keep that in mind. Good info!

Poly count is 21,350 with sub-d's set to 3.
Here's a test render I did last week before tweeking the lights and transparency maps a little.

Is that 21,350 polys before subdivision? If so the formula is:

# polys before sub-d * (sub-d level)^2 * 2

which makes only 384,300 polys... it's definately your maps.

But as I look at your image, I am wondering what parts of it actually need to be rendered? Since it's for a poster, it looks like you could build most of this right in Photoshop, and add a few rendered elements where you need them.

Unless I am somehow misunderstanding why you need it all rendered in LW...


CGTalk Moderation
01-16-2006, 08:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.